Conversational partners who repeatedly refer to the same objects require less and less collaborative effort to do so. This is due to lexical entrainment, the fact that they come to re-use the same words. Lexical entrainment may reflect the elaboration of conceptual pacts, partner-specific agreements about how to name objects which belong to the conversational partners’ common ground. Can lexical entrainment occur even if conversational partners cannot develop conceptual pacts about specific objects? In three experiments, we investigated whether lexical entrainment occurs in the matching task even when cards change over trials and partners are not able to develop pacts. We compared two conditions: a classic condition where cards remained the same for each trial, and a new cards condition where cards changed on each trial. Lexical diversity decreased for pairs in the new cards condition (albeit less than for classical pairs); inconsistent reductions in collaborative effort were also observed. Pairs in the new cards condition also were better able to adapt to novel referring situations (involving novel stimuli or new interaction partners) than classic pairs. The results suggest that lexical entrainment in the matching task may be due in part to factors other than the elaboration of conceptual pacts. These may include the development of an overarching meta-perspective on shared features of cards, reflecting category learning processes resulting from reference negotiation.